I want to use box ai to train my own ai models
I imagine a world where I can create a model and associate it with a metadata set. I think it would work like this.
Create New AI Model. I would put it right next to metadata creation. Let's say I create an AI model called "Analyst Answer". I would associate Analyst Answer with a metadata template.
In the AI Model I would pick the metadata template and then I would assign Independent and Dependent variables from the MetaData Template.
For example,
Independent Metadata Field -
Financial Input 1 (Net Worth) - Tag as Independent
Financial Input 2 (Income) - Tag as Independent
Financial Input 3 (Annual Performance) - Tag as Independent
Financial Input 4 (Age) - Tag as Independent
Output - (Analyst Commentary) - Tag as Dependent
I could handkey in those metadata values or use Box metadata extraction to fill them in.
I could do a search based on the files with that metadata and click select "train model".
Now, when I upload a new document of that metadata type it would run through the ai use metadata extraction to fill in the independent variables and then based on the model fill in the dependent or AI model value.
A way to leverage Box documents for models without having to pull things in and out.
